Eric Blue

The National Security Hackathon for Defense in San Francisco brought together over 500 participants across 64 teams, all tasked with addressing complex challenges in next-generation intelligence analysis, drones, computer vision, and AI, including GPT and LLMs. We focused on a significant problem: processing and extracting actionable intelligence from the overwhelming amount of unstructured data generated daily—approximately 300 million terabytes. Our solution aimed to transform this information overload into strategic insights by developing a sophisticated data pipeline.

Leveraging Large Language Models (LLMs) and a graph database, we built a system that extracted key entities from diverse data sources, such as unclassified military briefs, social media feeds, and global news outlets. These entities were used to construct a deep knowledge graph that provided insights into the relationships between people, places, events, and outcomes. Additionally, our solution featured capabilities for chat-based Q&A on the knowledge graph and summarization of intelligence over time.

Consulted through NGA’s Moonshot Labs (an unclassified innovation hub and collider space for NGA). Technical Program Manager (engineering, product, tech lead/architect) for a 12 person team successfully delivered 4 major projects/products over the course of 2+ years.

This included development of scalable architectures in support of geospatial data analysis, web-based tools for operations/executive use, and large-scale microservice-based architectures, APIs and web-based tools for mission stakeholders.

Agile team and product lead, however providing hands on assistance where needed for cloud migrations, security assessments and product launches

I wanted to see how quickly and effectively GPT-4V (gpt-4-1106-vision-preview) could analyze artwork given an uploaded image. I primarily wanted to do this for experimenting with my own artwork, but also interested in discovering what type of critique and feedback it would give on famous artist’s work.

This project was written in Python, uses Steamlit and integrates with OpenAI. I’d also done some early evals of LLaVa on HuggingFace.

In 2017, working through another lead consultancy (team size 15), I took on the role of technical lead (architect, BE developer). Harley-Davidson was looking to develop a new version of the web-based rideplanner app, and launch new iOS and Android apps.

Developed new codebase (Java/Spring Boot) and APIs for all backend systems

Migrated infrastructure originally from Pivotal Cloud, to PCF, and then ultimately spearheaded the migration to AWS/Kubernetes resulting in significant scaling and cost reduction

Original Technical Challenge - Offering social challenges, leaderboards, rewards. Evaluated multiple cloud-based solutions, and with a highly compressed timeframe ultimately designed and developed our own in house solutions - new social+leaderboard backend systems, APIs, admin tools and mobile integration

Design and built admin/operations/marketing support tools for location-based marketing and notifications

In mid-2016, Ian V.J. Murray, PhD, Associate Professor of Physiology and Neuroscience at St. George University approached ActiveOS with a funded research project that needed to steam, capture and analyze brainwave data utilizing a Neurosky EEG sensor.

No software on the market currently offered real-time streaming of concurrent EEG headsets via Bluetooth and no easy way to save this data to the cloud for performing historical storage and analysis of attention and various brainwave states.

In parallel with building the client technology, named MindStream, ActiveOS created an extension to the Enterprise Platform product for streaming and saving real-time EEG (thousands of records per second) and other biometric data.

• Added ability to simultaneously stream up to 6-8 EEG headsets using a custom bluetooth integration

• Integrated EEG data streaming using a combo local interface (windows, mac) and cloud-based streaming (ActiveoS+Vert.x)

Porsche + The Atlantic + ActiveOS (branded then as TRAQS) + Hexoskin to launch the new Porsche Macan in late 2014. ActiveOS powered the event with our Enterprise Platform, aggregating biometric and location data of car drivers and producing a highly-interactive touch screen dashboard on a 66” TV at the event. The video and featured story at The Atlantic was used for the Porsche Macan marketing campaign.

• Integrated with Hexoskin smart shirt, high-performance GPS trackers used by professional race car drivers

• Worked with Porsche and The Atlantic on the concept, design, development and ideation for the marketing campaign launch

• Live event in Chicago with 40+ notable personalities (exclusive invite) to drive Porsches and driven by professional race car drivers - visualizing heart rate, breathing and GPS data

• Data processed and streamed at the event on an interactive touch screen TV

• Official Atlantic + Porsche marketing video at https://vimeo.com/111218997

ActiveOS, a cloud-based wearable and geolocation data platform, partnering with global sensor companies and developing products for customers spanning the health & fitness industry, academia and research. This started out as a number of open source projects in the early days of wearables and morphed into a commercial product - our 1st major customer was Porsche in partnership with Hexoskin (smart shirt company)

This was my company - founder, ceo and primary developer for the majority of the platform

• Architected cloud-based architecture for wearable data integration

• Created plugin framework+SDK for all wearable integrations (OAuth2 - Fitbit, Garmin, Withings, etc.)

• Developed all backend systems and front-end apps (Java/Spring, Perl)

• Created a JavaScript-based framework for wearable data charts+visualizations

• Developed all FE Dashboards and team portals (BootStrap, JQuery, Node.js)